The Tennessee Department of Environment and Conservation lists sediment as the most commonly cited water pollutant in the state of Tennessee. In order to successfully implement management strategies to control sediment loss it is imperative to correctly identify the sources of the sediment. Without identifying sediment sources it is hard to pinpoint areas of highest priority, and also difficult to evaluate the effectiveness of management practices. The objective of this study is to evaluate a chemical fingerprinting technique for determining the source of in-stream sediment at the watershed scale. The Pond Creek watershed in east Tennessee was chosen for the study because it is representative of agricultural watersheds that have been significantly impacted by sediment in the ridge and valley physiographic region. Furthermore, Pond Creek and two of its tributaries, Greasy Branch and Mud Creek, are listed on the Tennessee 303(d) list of impaired streams for not satisfying all of their use requirements. Soil surface samples from the top two centimeters of the profile were collected from potential sediment sources within the watershed. Samples were also collected from sediment bars and the bottom of the stream. The silt plus clay fraction was isolated and subjected to total dissolution. The total content of 24 elements was determined using ICP. An array of multivariate statistical techniques was employed to analyze the elemental data. Principal component, hierarchical cluster, canonical discriminant analysis illustrated the tendency for stream sediments to cluster with potential sources. In general, sediments were derived from soils that were disturbed by dairy operations. Further, sources were spatially proximate to stream sediments. The results of this study suggest that the total elemental content of stream sediment and potential sources on the landscape may be employed to identify sediment sources at the watershed scale.